The Dartmouth Atlas of Health Care has demonstrated considerable variability among hospitals in intensive care unit (ICU) use at the end of life. Efforts to determine how much of this variability is due to differences in provider behavior are hindered by other factors that vary simultaneously across hospitals, such as patients'clinical condition, psychosocial circumstances, and preferences for treatment. We propose to augment our currently-funded NIH observational study of the medical intensive care units (MICUs) of a high-intensity academic medical center and a low-intensity academic medical center (R21 NR010265) with a simulation experiment at the same two institutions. While the observational study focuses on decisions that are made after patients have been admitted to the ICU, our simulation of a patient with end-stage cancer is designed to study the decision whether to admit a patient to the ICU in the first place, a decision event that is hard to study observationally, given the unpredictable timing of these events. In addition, by placing physicians in a simulated environment with an identical case, we can isolate the provider sources of variation among two hospitals at opposite ends of the end-of-life intensity spectrum. By exploring the rationales physicians offer to explain their behavior, we can further parse part of the causes of the variation into formal norms (hospital policies and procedures) and informal social norms. Of the many factors driving end-of-life decision making, informal social norms are one of the least understood. Social norms are also potentially modifiable with social marketing interventions. By complementing the data from our observational study with these experimental data, we will have a much clearer and broader picture of ICU use at the end of life, which we will use to inform a future intervention study to improve patient and family satisfaction with physician decision making. The specific aims are: Aim 1: To compare ICU admission, palliation, and code status documentation decisions among hospital- based physicians at one high-intensity and one low-intensity academic medical center using high-fidelity simulation. Aim 2: To examine the relationship between communication skills and ICU admission, palliation, and code status documentation decisions using simulation data from the 2 academic medical centers, augmented by previously collected data from a mid-intensity academic medical center. Aim 3: To identify formal and informal social norms that influence ICU admission, palliation, and code status documentation decisions at the high-intensity and the low-intensity academic medical centers. PUBLIC HEALTH RELEVANCE: The overall goal of this project is to enhance our understanding of the reasons for hospital-level variations in end-of-life (EOL) intensive care unit (ICU) use among patients with end-stage cancer. We will use a high- fidelity simulation, similar in sophistication to flight simulators used for pilots, to assess and compare the communication and decision-making processes of hospital-based physicians from an academic medical center with high EOL ICU use to those of physicians from an academic medical center with low EOL ICU use. By placing physicians in a simulated environment with an identical case, we can isolate the provider sources of variation among two hospitals at opposite ends of the end-of-life intensity spectrum. By exploring the rationales physicians offer to explain their behavior, we can further parse part of the causes of the variation into formal norms (hospital policies and procedures) and informal social norms. Our findings will be used to develop hospital-level interventions to improve the patient-centeredness of communication and decision making for dying patients. This study will also provide further support for the simulation method we have developed, which could be used in the future as a technique for identifying mechanisms underlying physician behavior and as a training tool for changing physician behavior.